Robots are rapidly becoming central tools to many industries, from manufacturing through e-commerce and autonomous vehicles to medicine and healthcare. A major challenge to current robotic manipulators is grasping objects, especially when the object shapes or poses are unknown in advance. Recent advances in Deep Learning have allowed training powerful grasp predictors, suggesting where a robot should attempt to grasp an object.
In this project, we will train an ensemble of grasp predictors to suggest grasp poses for novel objects, for a simulated robot environment. Using an ensemble of models instead of a single instance allows the robotic agent to reason about grasp uncertainty, and thus improve robustness of suggested grasps when using partial or noisy information about the object.
Note: this project can potentially be continued to a full-year project in multiple directions, such as moving from simulation to real robots.